Do not untrack resources of a server being unshelved

This patch concerns the time when a VM is being unshelved and the
compute manager set the task_state to spawning, claimed resources of
the VM and then called driver.spawn(). So the instance is in vm_state
SHELVED_OFFLOADED, task_state spawning.

If at this point a new update_available_resource periodic job is
started that collects all the instances assigned to the node to
calculate resource usage. However the calculation assumed that a
VM in SHELVED_OFFLOADED state does not need resource allocation on
the node (probably being removed from the node as it is offloaded)
and deleted the resource claim.

Given all this we ended up with the VM spawned successfully but having
lost the resource claim on the node.

This patch changes what we do in vm_state SHELVED_OFFLOADED, task_state
spawning. We no longer delete the resource claim in this state and
keep tracking the resource in stats.

Change-Id: I8c9944810c09d501a6d3f60f095d9817b756872d
Closes-Bug: #2025480
(cherry picked from commit f1dc4ec)
(cherry picked from commit 4239d1f)
(cherry picked from commit 683ecc0)
(cherry picked from commit e41962f)
(cherry picked from commit 22dfd1f)
(cherry picked from commit 26a9cae)

Author: rubasov

nova/compute/manager.py

@@ -6488,9 +6488,9 @@ def _shelve_offload_instance(self, context, instance, clean_shutdown,
 
         instance.power_state = current_power_state
         # NOTE(mriedem): The vm_state has to be set before updating the
-        # resource tracker, see vm_states.ALLOW_RESOURCE_REMOVAL. The host/node
-        # values cannot be nulled out until after updating the resource tracker
-        # though.
+        # resource tracker, see vm_states.allow_resource_removal(). The
+        # host/node values cannot be nulled out until after updating the
+        # resource tracker though.
         instance.vm_state = vm_states.SHELVED_OFFLOADED
         instance.task_state = None
         instance.save(expected_task_state=[task_states.SHELVING,

nova/compute/resource_tracker.py

@@ -1465,7 +1465,8 @@ def _update_usage_from_instance(self, context, instance, nodename,
         # NOTE(sfinucan): Both brand new instances as well as instances that
         # are being unshelved will have is_new_instance == True
         is_removed_instance = not is_new_instance and (is_removed or
-            instance['vm_state'] in vm_states.ALLOW_RESOURCE_REMOVAL)
+            vm_states.allow_resource_removal(
+                vm_state=instance['vm_state'], task_state=instance.task_state))
 
         if is_new_instance:
             self.tracked_instances.add(uuid)
@@ -1524,7 +1525,9 @@ def _update_usage_from_instances(self, context, instances, nodename):
 
         instance_by_uuid = {}
         for instance in instances:
-            if instance.vm_state not in vm_states.ALLOW_RESOURCE_REMOVAL:
+            if not vm_states.allow_resource_removal(
+                    vm_state=instance['vm_state'],
+                    task_state=instance.task_state):
                 self._update_usage_from_instance(context, instance, nodename)
             instance_by_uuid[instance.uuid] = instance
         return instance_by_uuid

nova/compute/stats.py

@@ -105,7 +105,8 @@ def update_stats_for_instance(self, instance, is_removed=False):
         (vm_state, task_state, os_type, project_id) = \
                 self._extract_state_from_instance(instance)
 
-        if is_removed or vm_state in vm_states.ALLOW_RESOURCE_REMOVAL:
+        if is_removed or vm_states.allow_resource_removal(
+                vm_state=vm_state, task_state=task_state):
             self._decrement("num_instances")
             self.states.pop(uuid)
         else:

nova/compute/vm_states.py

@@ -27,6 +27,7 @@
 See http://wiki.openstack.org/VMState
 """
 
+from nova.compute import task_states
 from nova.objects import fields
 
 
@@ -74,5 +75,11 @@
 # states we allow to trigger crash dump
 ALLOW_TRIGGER_CRASH_DUMP = [ACTIVE, PAUSED, RESCUED, RESIZED, ERROR]
 
-# states we allow resources to be freed in
-ALLOW_RESOURCE_REMOVAL = [DELETED, SHELVED_OFFLOADED]
+
+def allow_resource_removal(vm_state, task_state=None):
+    """(vm_state, task_state) combinations we allow resources to be freed in"""
+
+    return (
+        vm_state == DELETED or
+        vm_state == SHELVED_OFFLOADED and task_state != task_states.SPAWNING
+    )

nova/tests/functional/regressions/test_bug_2025480.py

@@ -77,6 +77,5 @@ def fake_spawn(*args, **kwargs):
 
         node = compute_node.ComputeNode.get_by_nodename(
             context.get_admin_context(), 'compute1')
-        # This is the bug, the instance should have resources claimed
-        # self.assertEqual(1, node.vcpus_used)
-        self.assertEqual(0, node.vcpus_used)
+        # After the fix, the instance should have resources claimed
+        self.assertEqual(1, node.vcpus_used)

nova/tests/unit/compute/test_stats.py

@@ -208,6 +208,22 @@ def test_update_stats_for_instance_offloaded(self):
         self.assertEqual(0, self.stats.num_os_type("Linux"))
         self.assertEqual(0, self.stats["num_vm_" + vm_states.BUILDING])
 
+    def test_update_stats_for_instance_being_unshelved(self):
+        instance = self._create_instance()
+        self.stats.update_stats_for_instance(instance)
+        self.assertEqual(1, self.stats.num_instances_for_project("1234"))
+
+        instance["vm_state"] = vm_states.SHELVED_OFFLOADED
+        instance["task_state"] = task_states.SPAWNING
+        self.stats.update_stats_for_instance(instance)
+
+        self.assertEqual(1, self.stats.num_instances)
+        self.assertEqual(1, self.stats.num_instances_for_project(1234))
+        self.assertEqual(1, self.stats["num_os_type_Linux"])
+        self.assertEqual(1, self.stats["num_vm_%s" %
+                                       vm_states.SHELVED_OFFLOADED])
+        self.assertEqual(1, self.stats["num_task_%s" % task_states.SPAWNING])
+
     def test_io_workload(self):
         vms = [vm_states.ACTIVE, vm_states.BUILDING, vm_states.PAUSED]
         tasks = [task_states.RESIZE_MIGRATING, task_states.REBUILDING,